Fri, 26 Mar 1993 09:58:37 GMT

Related articles |
---|

Effect of pointers etc. sharma@math.tu-berlin.de (1993-03-26) |

Re: Effect of pointers etc. preston@dawn.cs.rice.edu (1993-03-27) |

Re: Effect of pointers etc. paco@legia.cs.rice.edu (1993-03-28) |

Effect of pointers etc. ssimmons@convex.com (1993-03-29) |

Re: Effect of pointers etc. firth@sei.cmu.edu (1993-03-30) |

Newsgroups: | comp.compilers,comp.parallel |

From: | sharma@math.tu-berlin.de (Sharma) |

Keywords: | parallel, analysis, comment |

Organization: | FB3, TU-Berlin, Germany. |

Date: | Fri, 26 Mar 1993 09:58:37 GMT |

I am currently doing a part-time job in a Research Institute where I have

to determine whether an algorithm for 3DTV written sequentially in C is

implementable in real time on a parallel system. The final system could

be anything SIMD, Asynchronous, Vector Supercomputer or even dedicated

hardware. My job is to say how many operations are required. For example:

The module search_disparity requires x times 16 bit integer adds and y

times 32 bit floating point divides. The only way to go about it is to

plod my way through a source code listing as thick as a telephone book.

The algorithms contain lots of 2 dimensional arrays (naturally). Now what

is the cost in terms of hardware of macros such as these

#define DISP1(i,j) *(dispar.disp0 + i*MAX_X0 + j)

#define DISP(i,j) *(dispar.displr + i*MAX_X0 + j)

#define DISPold(i,j) *(dispar.disp_old + i*MAX_X0 + j)

Furthermore I would be thankful if anyone has any suggestions whatsoever

as to how to proceed with this mammoth task.

Thanks

Rohit Sharma

sharma@math.tu-berlin.de

[These macros are basically providing two-dimensional addressing in a clumsy

way probably because the person who wrote the code didn't know how to write

the declarations he wanted. But that doesn't help much with counting the

work involved. -John]

--

Post a followup to this message

Return to the
comp.compilers page.

Search the
comp.compilers archives again.